def instruction_process_engine(self):
self.init_conn()
ps = r.pubsub(ignore_subscribe_messages=False)
ps.subscribe(config['instruction_channel'])
while True:
if Utils.exit_flag:
msg = 'Thread instruction_process_engine say bye-bye'
print msg
logger.info(msg=msg)
return
threads_status['instruction_process_engine'] = dict()
threads_status['instruction_process_engine']['timestamp'] = ji.Common.ts()
# noinspection PyBroadException
try:
msg = ps.get_message(timeout=config['engine_cycle_interval'])
if msg is None or 'data' not in msg or not isinstance(msg['data'], basestring):
continue
try:
msg = json.loads(msg['data'])
if msg['action'] == 'pong':
continue
if msg['action'] == 'ping':
# 通过 ping pong 来刷存在感。因为经过实际测试发现，当订阅频道长时间没有数据来往，那么订阅者会被自动退出。
r.publish(config['instruction_channel'], message=json.dumps({'action': 'pong'}))
continue
except ValueError as e:
logger.error(e.message)
log_emit.error(e.message)
continue
if 'node_id' in msg and int(msg['node_id']) != self.node_id:
continue
# 下列语句繁琐写法如 <code>if 'action' not in msg or 'uuid' not in msg:</code>
if not all([key in msg for key in ['_object', 'action']]):
continue
extend_data = dict()
if msg['_object'] == 'guest':
self.refresh_guest_mapping()
if msg['action'] not in ['create']:
if msg['uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('The uuid ' + msg['uuid'] + ' not found in current domains list.')
self.guest = self.guest_mapping_by_uuid[msg['uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Guest ' + msg['uuid'] + ' is not a domain.')
if msg['action'] == 'create':
t = threading.Thread(target=Guest.create, args=(self.conn, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'reboot':
if self.guest.reboot() != 0:
raise RuntimeError('Guest reboot failure.')
elif msg['action'] == 'force_reboot':
self.guest.destroy()
self.guest.create()
Guest.quota(guest=self.guest, msg=msg)
elif msg['action'] == 'shutdown':
if self.guest.shutdown() != 0:
raise RuntimeError('Guest shutdown failure.')
elif msg['action'] == 'force_shutdown':
if self.guest.destroy() != 0:
raise RuntimeError('Guest force shutdown failure.')
elif msg['action'] == 'boot':
if not self.guest.isActive():
if self.guest.create() != 0:
raise RuntimeError('Guest boot failure.')
Guest.quota(guest=self.guest, msg=msg)
elif msg['action'] == 'suspend':
if self.guest.suspend() != 0:
raise RuntimeError('Guest suspend failure.')
elif msg['action'] == 'resume':
if self.guest.resume() != 0:
raise RuntimeError('Guest resume failure.')
elif msg['action'] == 'delete':
root = ET.fromstring(self.guest.XMLDesc())
if self.guest.isActive():
self.guest.destroy()
self.guest.undefine()
system_disk = None
for _disk in root.findall('devices/disk'):
if 'vda' == _disk.find('target').get('dev'):
system_disk = _disk
if msg['storage_mode'] in [StorageMode.ceph.value, StorageMode.glusterfs.value]:
# 签出系统镜像路径
path_list = system_disk.find('source').attrib['name'].split('/')
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = path_list[0]
Guest.init_gfapi()
try:
Guest.gf.remove('/'.join(path_list[1:]))
except OSError:
pass
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
file_path = system_disk.find('source').attrib['file']
try:
os.remove(file_path)
except OSError:
pass
elif msg['action'] == 'reset_password':
if self.guest.setUserPassword(msg['user'], msg['password']) != 0:
raise RuntimeError('Guest reset password failure.')
elif msg['action'] == 'attach_disk':
if 'xml' not in msg:
_log = u'添加磁盘缺少 xml 参数'
raise KeyError(_log)
flags = libvirt.VIR_DOMAIN_AFFECT_CONFIG
if self.guest.isActive():
flags |= libvirt.VIR_DOMAIN_AFFECT_LIVE
# 添加磁盘成功返回时，ret值为0。可参考 Linux 命令返回值规范？
if self.guest.attachDeviceFlags(xml=msg['xml'], flags=flags) != 0:
raise RuntimeError('Attack disk failure.')
Guest.quota(guest=self.guest, msg=msg)
elif msg['action'] == 'detach_disk':
if 'xml' not in msg:
_log = u'分离磁盘缺少 xml 参数'
raise KeyError(_log)
flags = libvirt.VIR_DOMAIN_AFFECT_CONFIG
if self.guest.isActive():
flags |= libvirt.VIR_DOMAIN_AFFECT_LIVE
if self.guest.detachDeviceFlags(xml=msg['xml'], flags=flags) != 0:
raise RuntimeError('Detach disk failure.')
elif msg['action'] == 'update_ssh_key':
if not self.guest.isActive():
_log = u'欲更新 SSH-KEY 的目标虚拟机未处于活动状态。'
logger.warning(_log)
log_emit.warn(_log)
continue
ret = Guest.update_ssh_key(guest=self.guest, msg=msg)
logger.info(json.dumps(ret, ensure_ascii=False))
elif msg['action'] == 'allocate_bandwidth':
t = threading.Thread(target=Guest.allocate_bandwidth, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'adjust_ability':
t = threading.Thread(target=Guest.adjust_ability, args=(self.conn, self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'migrate':
# duri like qemu+ssh://destination_host/system
if 'duri' not in msg:
_log = u'迁移操作缺少 duri 参数'
raise KeyError(_log)
# https://rk4n.github.io/2016/08/10/qemu-post-copy-and-auto-converge-features/
flags = libvirt.VIR_MIGRATE_PERSIST_DEST | \
libvirt.VIR_MIGRATE_UNDEFINE_SOURCE | \
libvirt.VIR_MIGRATE_COMPRESSED | \
libvirt.VIR_MIGRATE_PEER2PEER | \
libvirt.VIR_MIGRATE_AUTO_CONVERGE
root = ET.fromstring(self.guest.XMLDesc())
if msg['storage_mode'] == StorageMode.local.value:
# 需要把磁盘存放路径加入到两边宿主机的存储池中
# 不然将会报 no storage pool with matching target path '/opt/Images' 错误
flags |= libvirt.VIR_MIGRATE_NON_SHARED_DISK
flags |= libvirt.VIR_MIGRATE_LIVE
if not self.guest.isActive():
_log = u'非共享存储不支持离线迁移。'
logger.error(_log)
log_emit.error(_log)
raise RuntimeError('Nonsupport online migrate with storage of non sharing mode.')
if self.init_ssh_client(hostname=msg['duri'].split('/')[2], user='******'):
for _disk in root.findall('devices/disk'):
_file_path = _disk.find('source').get('file')
disk_info = Disk.disk_info_by_local(image_path=_file_path)
disk_size = disk_info['virtual-size']
stdin, stdout, stderr = self.ssh_client.exec_command(
' '.join(['qemu-img', 'create', '-f', 'qcow2', _file_path, str(disk_size)]))
for line in stdout:
logger.info(line)
log_emit.info(line)
for line in stderr:
logger.error(line)
log_emit.error(line)
elif msg['storage_mode'] in [StorageMode.shared_mount.value, StorageMode.ceph.value,
StorageMode.glusterfs.value]:
if self.guest.isActive():
flags |= libvirt.VIR_MIGRATE_LIVE
flags |= libvirt.VIR_MIGRATE_TUNNELLED
else:
flags |= libvirt.VIR_MIGRATE_OFFLINE
if self.guest.migrateToURI(duri=msg['duri'], flags=flags) == 0:
if msg['storage_mode'] == StorageMode.local.value:
for _disk in root.findall('devices/disk'):
_file_path = _disk.find('source').get('file')
if _file_path is not None:
os.remove(_file_path)
else:
raise RuntimeError('Unknown storage mode.')
elif msg['_object'] == 'disk':
if msg['action'] == 'create':
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
if not Disk.make_qemu_image_by_glusterfs(gf=Guest.gf, dfs_volume=msg['dfs_volume'],
image_path=msg['image_path'], size=msg['size']):
raise RuntimeError('Create disk failure with glusterfs.')
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
if not Disk.make_qemu_image_by_local(image_path=msg['image_path'], size=msg['size']):
raise RuntimeError('Create disk failure with local storage mode.')
elif msg['action'] == 'delete':
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
if Disk.delete_qemu_image_by_glusterfs(gf=Guest.gf, image_path=msg['image_path']) \
is not None:
raise RuntimeError('Delete disk failure with glusterfs.')
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
if Disk.delete_qemu_image_by_local(image_path=msg['image_path']) is not None:
raise RuntimeError('Delete disk failure with local storage mode.')
elif msg['action'] == 'resize':
if 'size' not in msg:
_log = u'添加磁盘缺少 disk 或 disk["size"] 参数'
raise KeyError(_log)
used = False
if msg['guest_uuid'].__len__() == 36:
used = True
if used:
self.refresh_guest_mapping()
if msg['guest_uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('Resize disk failure, because the uuid ' + msg['guest_uuid'] +
' not found in current domains.')
self.guest = self.guest_mapping_by_uuid[msg['guest_uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Resize disk failure, because the guest is not a domain.')
# 在线磁盘扩容
if used and self.guest.isActive():
if 'device_node' not in msg:
_log = u'添加磁盘缺少 disk 或 disk["device_node|size"] 参数'
raise KeyError(_log)
# 磁盘大小默认单位为KB，乘以两个 1024，使其单位达到GB
msg['size'] = int(msg['size']) * 1024 * 1024
if self.guest.blockResize(disk=msg['device_node'], size=msg['size']) != 0:
raise RuntimeError('Online resize disk failure in blockResize method.')
Guest.quota(guest=self.guest, msg=msg)
# 离线磁盘扩容
else:
if not all([key in msg for key in ['storage_mode', 'dfs_volume', 'image_path']]):
_log = u'添加磁盘缺少 disk 或 disk["storage_mode|dfs_volume|image_path|size"] 参数'
raise KeyError(_log)
if msg['storage_mode'] == StorageMode.glusterfs.value:
if not Disk.resize_qemu_image_by_glusterfs(dfs_volume=msg['dfs_volume'],
image_path=msg['image_path'],
size=msg['size']):
raise RuntimeError('Offline resize disk failure with glusterfs.')
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
if not Disk.resize_qemu_image_by_local(image_path=msg['image_path'], size=msg['size']):
raise RuntimeError('Offline resize disk failure with local storage mode.')
elif msg['action'] == 'quota':
self.refresh_guest_mapping()
if msg['guest_uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['guest_uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('Disk quota failure, because the uuid ' + msg['guest_uuid'] +
' not found in current domains.')
self.guest = self.guest_mapping_by_uuid[msg['guest_uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Disk quota failure, because the guest is not a domain.')
if not self.guest.isActive():
_log = u'磁盘 ' + msg['uuid'] + u' 所属虚拟机未处于活动状态。'
logger.warning(_log)
log_emit.warn(_log)
continue
Guest.quota(guest=self.guest, msg=msg)
elif msg['_object'] == 'snapshot':
self.refresh_guest_mapping()
if msg['uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('Snapshot ' + msg['action'] + ' failure, because the uuid ' +
msg['uuid'] + ' not found in current domains.')
self.guest = self.guest_mapping_by_uuid[msg['uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Snapshot ' + msg['action'] + ' failure, because the guest is not a domain.')
if msg['action'] == 'create':
t = threading.Thread(target=Guest.create_snapshot, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'delete':
t = threading.Thread(target=Guest.delete_snapshot, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'revert':
t = threading.Thread(target=Guest.revert_snapshot, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'convert':
t = threading.Thread(target=Guest.convert_snapshot, args=(msg,))
t.setDaemon(False)
t.start()
continue
elif msg['_object'] == 'os_template_image':
if msg['action'] == 'delete':
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
try:
Guest.gf.remove(msg['template_path'])
except OSError:
pass
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
try:
os.remove(msg['template_path'])
except OSError:
pass
elif msg['_object'] == 'global':
if msg['action'] == 'refresh_guest_state':
host_use_for_refresh_guest_state = Host()
t = threading.Thread(target=host_use_for_refresh_guest_state.refresh_guest_state, args=())
t.setDaemon(False)
t.start()
continue
else:
_log = u'未支持的 _object：' + msg['_object']
logger.error(_log)
log_emit.error(_log)
response_emit.success(_object=msg['_object'], action=msg['action'], uuid=msg['uuid'],
data=extend_data, passback_parameters=msg.get('passback_parameters'))
except redis.exceptions.ConnectionError as e:
logger.error(traceback.format_exc())
# 防止循环线程，在redis连接断开时，混水写入日志
time.sleep(5)
except:
# 防止循环线程，在redis连接断开时，混水写入日志
time.sleep(5)
logger.error(traceback.format_exc())
log_emit.error(traceback.format_exc())
response_emit.failure(_object=msg['_object'], action=msg.get('action'), uuid=msg.get('uuid'),
passback_parameters=msg.get('passback_parameters'))
def create(cls, conn, msg):
try:
Guest.storage_mode = msg['storage_mode']
guest = Guest(uuid=msg['uuid'],
name=msg['name'],
template_path=msg['template_path'],
disk=msg['disks'][0],
xml=msg['xml'])
if Guest.storage_mode == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
guest.system_image_path = guest.disk['path']
q_creating_guest.put({
'storage_mode': Guest.storage_mode,
'dfs_volume': Guest.dfs_volume,
'uuid': guest.uuid,
'template_path': guest.template_path,
'system_image_path': guest.system_image_path
})
if not guest.generate_system_image():
raise RuntimeError('System image generate failure.')
if not guest.define_by_xml(conn=conn):
raise RuntimeError(
'Define the instance of virtual machine by xml failure.')
guest_event_emit.creating(uuid=guest.uuid, progress=92)
disk_info = dict()
if Guest.storage_mode == StorageMode.glusterfs.value:
disk_info = Disk.disk_info_by_glusterfs(
dfs_volume=guest.dfs_volume,
image_path=guest.system_image_path)
elif Guest.storage_mode in [
StorageMode.local.value, StorageMode.shared_mount.value
]:
disk_info = Disk.disk_info_by_local(
image_path=guest.system_image_path)
# 由该线程最顶层的异常捕获机制，处理其抛出的异常
guest.execute_os_template_initialize_operates(
guest=conn.lookupByUUIDString(uuidstr=guest.uuid),
os_template_initialize_operates=msg[
'os_template_initialize_operates'],
os_type=msg['os_type'])
extend_data = dict()
extend_data.update({'disk_info': disk_info})
guest_event_emit.creating(uuid=guest.uuid, progress=97)
if not guest.start_by_uuid(conn=conn):
raise RuntimeError(
'Start the instance of virtual machine by uuid failure.')
cls.quota(guest=conn.lookupByUUIDString(uuidstr=guest.uuid),
msg=msg)
response_emit.success(
_object=msg['_object'],
action=msg['action'],
uuid=msg['uuid'],
data=extend_data,
passback_parameters=msg.get('passback_parameters'))
except:
logger.error(traceback.format_exc())
log_emit.error(traceback.format_exc())
response_emit.failure(
_object=msg['_object'],
action=msg.get('action'),
uuid=msg.get('uuid'),
passback_parameters=msg.get('passback_parameters'))